IMHO/IMHU AI can't claim authorship and as such can't copyright their work.

This doesn't prevent any form of automatic copyrighting by production of derivative code or similar. It just prevent anyone from claiming ownership of any parts unique to the derived work.

Like think about it if a natural disaster changes (e.g. water damages) a picture you did draw then a) you can't claim ownership of the natural produced changes but b) still have ownership of the original picture contained in the changed/derived work.

AI shouldn't change that.

Which brings us to another 2 aspects:

1. if you give an AI a project access to the code to rewrite it anew it _is_ a copyright violation as it's basically a side-by-side rewrite

2. but if you go the clean room approach but powered by AI then it likely isn't a copyright violation, but also now part of the public domain, i.e. not yours

So yes, doing clean room rewrites has become incredible cheap.

But no just because it's AI it doesn't make code go away.

And lets be realistic one of the most relevant parts of many open source project is it being openly/shared maintained. You don't get this with clean room rewrites no matter if AI or not.

LLM are isomorphic plagiarism machines, and like all ectoparasites must steal from real people to exist. Note this includes its users. =3